The Plant Cell, Vol. 15, 2003–2019, September 2003, www.plantcell.org © 2003 American Society of Plant Biologists
Domain Analysis of the Chloroplast Polynucleotide
Phosphorylase Reveals Discrete Functions in RNA
Degradation, Polyadenylation, and Sequence Homology
with Exosome Proteins
Shlomit Yehudai-Resheff,a Victoria Portnoy,a Sivan Yogev,a Noam Adir,b and Gadi Schuster a,1
a Department
b Department
of Biology, Technion–Israel Institute of Technology, Haifa 32000, Israel
of Chemistry, Technion–Israel Institute of Technology, Haifa 32000, Israel
The molecular mechanism of mRNA degradation in the chloroplast consists of sequential events, including endonucleolytic
cleavage, the addition of poly(A)-rich sequences to the endonucleolytic cleavage products, and exonucleolytic degradation.
In spinach chloroplasts, the latter two steps of polyadenylation and exonucleolytic degradation are performed by the same
phosphorolytic and processive enzyme, polynucleotide phosphorylase (PNPase). An analysis of its amino acid sequence
shows that the protein is composed of two core domains related to RNase PH, two RNA binding domains (KH and S1), and
an -helical domain. The amino acid sequence and domain structure is largely conserved between bacteria and organelles.
To define the molecular mechanism that controls the two opposite activities of this protein in the chloroplast, the ribonuclease, polymerase, and RNA binding properties of each domain were analyzed. The first core domain, which was predicted
to be inactive in the bacterial enzymes, was active in RNA degradation but not in polymerization. Surprisingly, the second
core domain was found to be active in degrading polyadenylated RNA only, suggesting that nonpolyadenylated molecules
can be degraded only if tails are added, apparently by the same protein. The poly(A) high-binding-affinity site was localized
to the S1 domain. The complete spinach chloroplast PNPase, as well as versions containing the core domains, complemented the cold sensitivity of an Escherichia coli PNPase-less mutant. Phylogenetic analyses of the two core domains
showed that the two domains separated very early, resulting in the evolution of the bacterial and organelle PNPases and the
exosome proteins found in eukaryotes and some archaea.
INTRODUCTION
Polynucleotide phosphorylase (PNPase) is an exoribonuclease
that catalyzes the processive 3 → 5 phosphorolysis of RNA
or a processive polymerization of this molecule ( Littauer and
Grunberg-Manago, 1999). In Escherichia coli, this enzyme is
mostly active in 3 → 5 phosphorolysis during RNA degradation or processing (Grunberg-Manago, 1999; Jarrige et al., 2002).
Limited polymerization activity can be detected in vivo under
certain growth conditions or when the gene for the poly(A)polymerase (PAP) is inactivated (Sarkar, 1997; Mohanty and
Kushner, 2000). In E. coli, a small proportion of PNPase is a constituent of the degradosome, a multiprotein complex composed
also of RNase E, RNA helicase, enolase, and possibly other molecules (Symmons et al., 2002). The association of PNPase with
the RNA helicase RhlB alone also was observed recently ( Liou
et al., 2002). However, PNPase in the chloroplast was found to
form a homotrimeric complex eluting from a size-exclusion column at 600 kD and lacks any known interactions with other
proteins (Baginsky et al., 2001). PNPase also was reported to
be a global regulator of virulence and persistence in Salmonella
1 To
whom correspondence should be addressed. E-mail gadis@
tx.technion.ac.il; fax 972-4-8295587.
Article, publication date, and citation information can be found at
www.plantcell.org/cgi/doi/10.1105/tpc.013326.
enterica (Clements et al., 2002). Recently, the human PNPase
was identified in an overlapping-pathway screen to discover genes
that displayed coordinated expression as a consequence of the
terminal differentiation and cellular senescence of human melanoma cells (Leszczyniecka et al., 2002).
The amino acid sequences of different PNPases from bacteria, as well as from the nuclear genomes of plants, yeast, and
mammals, display a high level of identity and feature similar structures composed of five motifs (Symmons et al., 2000, 2002; Zuo
and Deutscher, 2001; Raijmakers et al., 2002). These consist of
two core domains having different degrees of identity to the E.
coli phosphorylase RNase PH, an -helical domain between
the two core domains, and two adjacent RNA binding domains
(KH and S1) that also are found in other RNA binding proteins.
X-ray crystallographic analysis was used to reveal the threedimensional structure of the PNPase from the bacterium Streptomyces antibioticus. The enzyme is arranged in a homotrimeric
complex forming a “doughnut” shape surrounding a central channel that could accommodate a single-stranded RNA molecule
(Symmons et al., 2000, 2002). Similar structure and homology of
the two core domains were assigned recently to the exosome, a
multiprotein complex that functions in the 3 → 5 degradation
of RNA in the cytoplasm and nucleus of eukaryotic cells (Aloy
et al., 2002; Raijmakers et al., 2002).
The molecular mechanism of RNA degradation in the chloroplast has been elucidated and was found to be very similar to
2004
The Plant Cell
that of bacteria (Hayes et al., 1999; Schuster et al., 1999; Monde
et al., 2000). In both bacteria and chloroplasts, the first event is
endonucleolytic cleavage of the RNA molecule, followed by the
addition of a poly(A) tail in bacteria and a poly(A) tail (Komine et
al., 2000) or a poly(A)-rich tail (Lisitsky et al., 1996) in chloroplasts. The polyadenylated cleavage products then are directed
to rapid exonucleolytic degradation by PNPase and RNase II in
E. coli and by PNPase and possibly other exoribonucleases in
the chloroplast (Lisitsky et al., 1997a, 1997b; Lisitsky and Schuster,
1999). Therefore, polyadenylation is part of the RNA degradation mechanism in bacteria, in chloroplasts, and possibly also in
plant mitochondria (Carpousis et al., 1999; Coburn and Mackie,
1999; Gagliardi and Leaver, 1999; Hayes et al., 1999; Lupold et
al., 1999; Schuster et al., 1999; Monde et al., 2000; Regnier and
Arraiano, 2000; Kuhn et al., 2001).
As discussed above, polyadenylation in E. coli is performed
mostly by PAP, with PNPase functioning as a polymerase only
in the absence of PAP (Mohanty and Kushner, 2000). However,
we recently found that no PAP can be detected in spinach
chloroplasts, and thus both polyadenylation and degradation
are performed by one enzyme, PNPase (Yehudai-Resheff et al.,
2001). A similar situation also was found recently in cyanobacteria (Rott et al., 2003). To understand how the spinach chloroplast PNPase performs these opposing activities of polymerization
and phosphorolysis, and because the activity of the full-length
protein is mediated by the sum of activities of the different domains, we decided to analyze the different domains for polymerization, degradation, and RNA binding properties. We found that
the core domains had distinct activities and that their cooperation
was required to achieve their functionality.
RESULTS
The Spinach Chloroplast PNPase Structure Is Similar
to That of the Bacterial Enzyme
The spinach PNPase is homologous with the bacterial enzyme,
with an additional N-terminal transit peptide of 61 amino acids
and a 22–amino acid C-terminal extension (Figure 1, top). The
transit peptide is cleaved off during translocation of the preprotein to the chloroplast; therefore, it is not present in the mature
protein. Like the bacterial PNPases, the protein is composed of
two core domains, the second of which is homologous with the
RNase PH, another bacterial exoribonuclease that is a phosphorylase (first [1st] and second [2nd] cores in Figure 1). The
two core domains share some degree of identity; therefore, low
similarity between RNase PH and the 1st core domain also is observed (Symmons et al., 2002). The 1st core domain is followed by
an -helical domain, and the 2nd core domain is followed by the
KH and S1 domains, which are predicted to function in RNA binding and are present in many RNA binding proteins (Symmons et
al., 2000, 2002; Zuo and Deutscher, 2001) (Figure 1).
Because the amino acid sequence of the spinach chloroplast
PNPase shares a high degree of identity with that of the S. antibioticus enzyme, we used the coordinates of the bacterial enzyme to build a putative structure of the spinach PNPase. Figures 1A and 1B present the predicted monomeric structure of
the chloroplast PNPase compared with the S. antibioticus en-
zyme. The structures are very similar with respect to the relative
positions of the various subdomains. Both the 1st and 2nd core
domains provide the -sheet strands, which function as the trimerization interfaces, and the vicinity of the phosphorolytic catalytic site (binding of the tungsten) known for the 2nd domain
(Symmons et al., 2000, 2002). The -helical domain is situated
separately from the main structure of the 1st and 2nd core domains, and although it is arranged similarly, it is not identical for
the bacterial and chloroplast proteins. The KH and S1 domains,
as well as the extra amino acids at the C terminus of the chloroplast enzyme, also are separated from the core domains.
The arrangement of three PNPase polypeptides in the doughnut-shaped trimer is shown in Figures 1C and 1D. The formation of the trimers creates a central channel that is of the correct dimensions for a single-stranded RNA molecule (Symmons
et al., 2000, 2002). The KH and S1 domains of the three subunits all point in one direction, suggesting the possibility of binding the RNA molecule before it enters the channel and undergoes degradation there. Indeed, the spinach chloroplast PNPase
was purified as a homomultimer that fractionated on a sizeexclusion column at 600 kD (Baginsky et al., 2001). It is not known
whether the PNPase is a trimer that is eluted at this size or
whether two trimers join to form a hexamer.
Preparation of Recombinant PNPase and the
Different Fragments
The molecular analysis of RNA polyadenylation and degradation in spinach chloroplasts revealed that both activities are
performed by PNPase (Yehudai-Resheff et al., 2001). To better
understand how the enzyme activity is shifted to favor polymerization or degradation, we analyzed the activity of the protein
and its component parts in detail. To do so, we prepared the
PNPase recombinant protein lacking the chloroplast transit peptide as described in Methods, as well as several deletion proteins containing different domains (Figure 1E). The recombinant
proteins were purified and analyzed by SDS-PAGE to determine
their apparent molecular mass and purity (Figure 1F). In addition,
all proteins were verified by immunoblot analysis using antibodies against either the PNPase or the His6 tag (data not shown).
RNA Degradation and Polyadenylation Activities of the
Spinach Chloroplast PNPase and Its Domains
As discussed above, structural analysis of the S. antibioticus
enzyme predicted that only the 2nd core would be active in
phosphorolysis. This conclusion was strengthened recently by
mutational analysis of the E. coli PNPase, in which most of the
mutations eliminating the phosphorolysis activity were located
in the 2nd core domain in the vicinity of the tungsten binding
site in the S. antibioticus PNPase (Jarrige et al., 2002). Here, we
characterized the chloroplast PNPase and its fragmented domains for RNA degradation and phosphorolysis activities. The
degradation activity was assayed by incubating the corresponding protein with 32P-labeled RNA in the presence of 10 mM Pi
(Figure 2A). Indeed, the full-length (FL; data not shown) and
FL-S1 proteins rapidly degraded the RNA (Figure 2A). Interestingly, even though the RNA substrate used here was chosen
Domain Analysis of PNPase
2005
Figure 1. Molecular Homology-Based Model of the Spinach Chloroplast PNPase, and Protein Constructs Used in This Work.
(A) to (D) The domain structures of the bacterial and chloroplast PNPases are presented schematically at top. The boxes represent the different domains, as indicated above. TP indicates the chloroplast transit peptide. The C-terminal 22–amino acid extension is unique to the chloroplast PNPase.
The resolved structures ([A] and [C]) and predicted models ([B] and [D]) of the PNPase enzymes from S. antibioticus ([A] and [C]) and spinach chloroplast ([B] and [D]) are shown. The monomer structure is shown in (A) and (B), and the trimer structure is shown in (C) and (D). The different domains
and the trimerization interfaces are indicated in the spinach chloroplast PNPase (B). The circle indicates the location of the phosphorylase activity site
at the 2nd core domain as identified from tungsten binding of the S. antibioticus enzyme (B). The trimeric doughnut-like structure is presented in (C)
and (D) in an orientation allowing easy observation of the middle channel. Each monomer is colored differently, and in one monomer of the spinach
enzyme each domain is colored as outlined in the scheme presented at top. The homology-based modeling was performed using the 3D-PSSM Fold
Recognition Server at http://www.sbg.bio.ic.ac.uk/~3dpssm/. The complex was built by applying the crystal symmetry of the structure using the
Quanta program (Accelrys). The figure was constructed using Insight II (Accelrys).
(E) Scheme of the spinach PNPase protein. The full-length protein (FL) was produced in bacteria without the addition of a His6 tag (see text). The other
versions were expressed in E. coli fused to the His6 tag at the C terminus, as shown by the hatched boxes.
(F) Silver-stained polyacrylamide gel profile of the recombinant proteins (35 to 50 ng) after expression in an E. coli strain lacking the endogenous PNPase. Proteins containing the His6 tag (lanes 2 to 7) were purified by affinity chromatography and anion-exchange (MonoQ) steps and loaded onto a
10% SDS-PAGE gel. The FL protein (lane 1) was purified biochemically on size-exclusion, heparin, and anion-exchange columns. Molecular mass
markers are shown at left.
2006
The Plant Cell
Figure 2. Polymerization and Degradation Activities of the Different Proteins.
(A) to (C) Synthetically transcribed 32P-RNA corresponding to part of the chloroplast gene petD (petD-Dra) was incubated with the proteins and 10
mM Pi to determine the presence of degradation activity (A). The Pi was replaced with 1 mM ADP or GDP to measure the polyadenylation and
polyguanylation activities ([B] and [C], respectively). Samples were withdrawn at 0, 15, 35, 60, and 90 min, and the RNA was analyzed by denaturing
PAGE and autoradiography. The proteins used are indicated at top, and the input RNA, as well as the polymerized and degradation products, are indicated at right.
(D) Thin layer chromatography (TLC) analysis of the degradation products. Proteins were incubated with uniformly 32P-UTP–labeled RNA. After the incubation, the reaction products were spotted onto a polyethyleneimine TLC plate, which was developed with LiCl, dried, and autoradiographed. Lane
1, no protein; lane 2, FL-S1; lane 3, 2ndKHS1; lane 4, 1stH; lane 5, E. coli PNPase; lane 6, KHS1. Monophosphate, diphosphate, and triphosphate nucleotides of A, G, C, and U were analyzed on the same TLC plate and visualized by fluorescence quenching (lanes M). 32P-Pi also was separated on this plate as a marker. The migration patterns of the markers are indicated with circles.
Domain Analysis of PNPase
because it does not contain a predicted stem-loop structure,
the enzyme paused at a certain sequence and a degradation
product accumulated. No differences were observed when the
activity of the recombinant FL protein was compared with that
of purified PNPase from spinach chloroplasts (Yehudai-Resheff
et al., 2001) (data not shown). The 1st H protein, composed of
the 1st core and the -helical domains, also was active in RNA
degradation ( Figure 2A). This result was surprising, because
previous data did not predict activity, given the tungsten binding
data and the mutagenesis of the bacterial PNPases ( Symmons
et al., 2000; Jarrige et al., 2002). Therefore, the chloroplast enzyme may differ from the bacterial PNPases in the activity of the
1st core domain.
Even more surprising was the observation that the 2nd core
domain, which was predicted to harbor the active site either
with or without the KH and S1 domains, displayed very low RNA
degradation activity even in the presence of 10 mM Pi (Figure
2A). Only a small amount of the substrate RNA was digested,
compared with that in the 1st core and the FL proteins. However, this issue was resolved when RNA polyadenylation was
assayed in the presence of ADP without the addition of Pi.
Under these conditions, the 2nd core displayed both polymerization and RNA degradation activities ( Figure 2B). The polyadenylation activity of the 2nd core was transient and preceded
degradation, as observed previously for the FL protein (YehudaiResheff et al., 2001). No polyadenylation activity was obtained
under these conditions with the 1st core (Figure 2B). These results suggested that the degradation activity of the 2nd core is
dependent on previous polyadenylation, again similar to what
we observed in the lysed chloroplast system (Yehudai-Resheff
et al., 2001).
In addition, there also is the possibility that polyadenylation
preceding degradation occurs with the 1st core but is too rapid
or highly transient and thus not detected here. To analyze whether
or not the 1st domain expresses polymerization activity, the polymerization assay was repeated with GDP replacing ADP. Under
these conditions, the RNA is polyguanylated by PNPase, and
because poly(G) forms a strong tertiary structure that efficiently
inhibits the exonuclease activity, only the polymerization activity
is observed (Sundquist, 1993; Yehudai-Resheff et al., 2001). In
addition, such an assay enabled us to determine whether the
polymerization activity of the 2nd core is required for the subsequent degradation activity, because if it is, only polymerization would be obtained with GDP. Indeed, when the polyguanylation assay was performed, polymerization activities were
obtained with the FL-S1, 2ndKHS1, and 2nd domains, and
RNA degradation activity was inhibited (Figure 2C). No polymerization activity was observed with the 1st H protein, suggesting that the 1st domain is active only in RNA degradation
and not in polymerization. Because the RNA degradation and
polymerization activities of the 2nd KHS1 and the 2nd domains alone were very similar, we concluded that, as predicted,
these activities were located at the 2nd core domain.
To analyze the degradation products generated by the exoribonuclease activities of the two domains, thin layer chromatography was performed (Figure 2D). As expected from a phosphorylase, the degradation activities of all proteins examined
resulted in the formation of nucleoside diphosphates. There-
2007
fore, the product detected with 32P–UTP-RNA was 32P-UDP
and that detected with 32P–ATP-RNA was 32P-ADP (Figure 2D).
Together, the results of the experiments shown in Figure 2
demonstrate that for the spinach chloroplast PNPase, the 1st
core domain is active in RNA degradation but not in polymerization, whereas the 2nd core domain is active in polyadenylationdependent RNA degradation.
The High-Affinity Poly(A) Binding Site Is Located in the
S1 Domain
The bacterial and spinach PNPase proteins were characterized
previously using the UV light cross-linking assay as RNA binding proteins (Lisitsky et al., 1997b; Lisitsky and Schuster, 1999).
In these experiments, high affinity for poly(A) and poly(U) was
observed, suggesting an explanation for how this enzyme competes in bacteria or within the chloroplast for polyadenylated
RNA over nonpolyadenylated RNA. We wished to determine
whether this protein contains more than one RNA binding site
and in which domain(s) high-affinity poly(A) binding is located.
First, RNA binding was tested by UV light cross-linking, in which
the protein is incubated with 32P-RNA followed by UV irradiation, digestion of the RNA with ribonucleases, and analysis by
PAGE and autoradiography. Figure 3A presents the results of
this experiment using an RNA corresponding to the chloroplast
psbA gene. All of the proteins, except for the one composed of
only the 2nd core domain, bound RNA. This result suggested
that the 2nd core domain, although harboring a phosphorolysis
active site, does not bind RNA in such a way that it could be
detected by the UV light cross-linking assay used here. However, the 1stH protein bound RNA. As expected, the protein
composed of the KH and S1 domains, which were predicted to
be RNA binding domains, bound RNA (Figure 3A).
Next, we wanted to locate the high-affinity site for poly(A)
binding. To this end, the UV light cross-linking competition assay was used. In this assay, the signal obtained by the UV light
cross-linking of 32P-RNA to protein is competed by adding increasing amounts of the tested nonradioactive RNA, in this
case, ribohomopolymers. The efficiency with which an RNA
competes for UV light cross-linking reflects its affinity for the
protein. The IC50 was defined as the molar excess of the competitor RNA that resulted in a 50% reduction in the radioactive
UV light cross-linking signal ( Lisitsky et al., 1994). The lower
the IC50 value for a specific RNA, the higher its affinity for the
protein. An example of such a UV light competition assay is
presented in Figure 3B, and quantification of RNA binding to
the different proteins is shown in Figures 3C and 3D. A high
binding affinity of the FL protein was observed for both poly(A)
and poly(U), as reported previously for the purified protein
(Lisitsky et al., 1997b). Interestingly, all of the proteins (except
the 2nd core, which did not bind RNA in the UV light cross-linking assay and therefore was not analyzed here) bound poly(U)
with high affinity. However, high affinity for poly(A) was observed only with proteins containing the S1 domain. The affinity
for poly(A) was reduced sixfold compared with the FL protein
when only the S1 domain was eliminated, and it was reduced
ninefold compared with the 1stH protein. Together, the results of the RNA binding analysis suggest that the poly(A) high-
2008
The Plant Cell
Figure 3. RNA Binding Properties of the Derivative Proteins.
(A) The proteins (10 ng each) FL (lane 1), FL-S1 (lane 2), 1stH2nd (lane 3), 1stH (lane 4), 2ndKHS1 (lane 5), 2nd (lane 6), and KHS1 (lane 7) (as
shown in Figure 1) were analyzed for RNA binding by the UV light cross-linking assay as described in Methods.
(B) Competition of ribohomopolymers for RNA binding of the KHS1 protein (protein 7 in Figure 1) was performed using 32P-psbA RNA in a UV light
cross-linking competition assay. The numbers at top indicate the molar excess of the ribohomopolymer.
(C) Competition of ribohomopolymers for RNA binding of the 1stH and 2ndKHS1 proteins (proteins 4 and 5 in Figure 1, respectively). The UV light
cross-linking competition assays were performed as described for (B), and the intensity of the UV light cross-linking band without competitor was defined as 100%. The data shown are averages of at least three independent experiments. IC50 values (competitor excess that resulted in 50% inhibition
of the UV light cross-linking signal) are indicated by dashed lines.
(D) UV light cross-linking competition assays were performed as described for (B) and (C). IC50 values for binding the ribohomopolymers poly(A) and
poly(U) by the different proteins are presented.
binding-affinity site is located in the S1 domain and that other RNA
binding sites of the PNPase possess a high affinity for poly(U).
The Spinach Chloroplast PNPase and Its Active Fragments
Complement the Growth of an E. coli PNPase- and RNase
PH–less Strain at 18C
The E. coli strain SK 8992 contains an insertion of the Tn 5
transposable element into the pnp gene encoding the PNPase
and also lacks the other Pi-dependent exoribonuclease RNase
PH. Although this strain lacks the PNPase, it is viable, probably
because RNase II can compensate for the PNPase RNA degradation activity. However, this strain is sensitive to cold and cannot grow at 18C (Yancey and Kushner, 1990; Craven et al., 1992;
Zhou and Deutscher, 1997; Beran and Simons, 2001). We used
SK 8992 to reveal whether the spinach chloroplast PNPase could
complement its E. coli counterpart and, if so, which domains of
the protein also could do so. To this end, the FL spinach chloroplast PNPase and its derivatives were cloned into the PT7-7
vector (Citovsky et al., 1990), which enables the expression of
recombinant proteins without the addition of other amino acids.
The plasmids then were introduced into SK 8992 cells. SK 8992
is PNPase- and RNase PH–less but does not contain the T7
RNA polymerase encoded in the chromosome. We had to shift
to this strain because the ENS134 strain containing the T7 RNA
polymerase used for the expression of the recombinant protein
grew at 18C when the PT7-7 plasmid alone, which does not express any part of the PNPase, was introduced (data not shown).
As described above, shifting to SK 8992, which does not contain the T7 RNA polymerase gene, solved this problem.
Taking this into account, and because it remained unclear
how the expression of the recombinant proteins was established,
their expression was verified by immunoblot analysis of proteins
from bacteria grown at 18C (Figure 4, right). As expected, the SK
8992 cells grew well at 37C but not at 18C, compared with the
control cells, which were transformed with a plasmid expressing the E. coli PNPase (Figure 4, top two rows). Interestingly,
full complementation was obtained with the FL protein, indicating that the chloroplast PNPase can efficiently restore the growth
defect at 18C of the E. coli strain lacking PNPase and RNase
PH (Figure 4, third row). In addition, all of the deletion mutants of
the chloroplast PNPase containing at least one core domain
were able to partially complement the cold growth defect. The
proteins containing only one core were less efficient than those
Domain Analysis of PNPase
2009
Figure 4. Complementation of Growth of the E. coli pnp Strain at 18C by the Chloroplast PNPase.
SK 8992 cells were transformed with a plasmid expressing the E. coli enzyme as a positive control (E. coli PNPase), the plasmid vector PT7-7 [SK
8992 (pnp)], and the different derivatives of the spinach chloroplast PNPase as indicated. This strain also lacks the other Pi-dependent exoribonuclease, RNase PH. The cells were grown overnight at 37C and then spotted onto L-agar/ampicillin plates at the dilutions shown at bottom. The plates
were incubated for 16 and 48 h at 37 and 18C, respectively. To verify expression of the corresponding proteins, the bacteria also were grown in cultures at 18C and analyzed for recombinant protein expression by immunoblotting using the PNPase antibodies (gels at right).
containing both cores, but even the protein composed of only
the 2nd core domain partially complemented the temperaturesensitive growth defect. These results support the biochemical
assay showing that the 1st core domain of the chloroplast PNPase is active in RNA degradation, an activity that probably is required to restore growth at 18C (Zhou and Deutscher, 1997).
Although the biochemical data suggested that this domain
does not express polyadenylation activity, polyadenylation could
be performed in E. coli by PAP I. RNase PH alone (containing
only one core domain) was shown previously to complement the
growth of the double mutant (PNPase and RNase PH) at 31C
(Zhou and Deutscher, 1997) but not at 15C (Beran and Simons,
2001). Therefore, although it seems unlikely given that 18C was
used in our experiments, the possibility cannot be excluded that
each of the spinach PNPase core domains alone actually complemented the RNase PH and not the PNPase function and that
this complementation enabled slow growth at 18C.
In contrast to the proteins containing the core domains, expression of the KHS1 protein did not restore growth at 18C (Figure
4, bottom row). Moreover, the expression of this protein inhibited
bacterial growth even at 37C, suggesting that the RNA binding
properties of this protein, when not connected to the core domains, are deleterious to the bacteria.
Together, these results showed that the chloroplast PNPase
could fully complement the cold defect of E. coli lacking PNPase
and RNase PH. Each part containing one of the core domains can
partially complement this defect, but the protein composed of only
the KHS1 domain inhibits growth even at 37C.
Unlike the FL PNPase, the Proteins That Include Only One
Core Domain Do Not Pause at a Stem-Loop Structure
One of the well-known characteristics of the PNPase enzyme is
its pausing at a stem-loop structure when processively degrading
RNA (Hayes et al., 1996; Blum et al., 1999; Liou et al., 2002). This
phenomenon is very important for the 3 end processing of bacterial and chloroplast transcripts, because the 3 end of most transcripts is characterized by a stable stem-loop structure formed by
the exonucleolytic trimming of a longer precursor (Carpousis et
al., 1999; Monde et al., 2000). Because homotrimer formation is
dependent on the interfaces formed by the two core domains
(Figure 1), no trimers are formed in the absence of one of them.
Indeed, evidence that each of the domains alone could not form
a high-molecular-weight complex was obtained experimentally
when the recombinant proteins were fractionated by size-exclusion chromatography or nondenaturing polyacrylamide gels (data
not shown). Because both core domains were found to be active
in RNA degradation, we asked whether each core, which probably
does not form the trimeric doughnut/channel conformation, pauses
at the stem-loop structure, like the FL enzyme.
To answer this question, an RNA molecule corresponding to
the 3 end of the spinach chloroplast psbA transcript and containing a stem-loop structure was incubated with the FL enzyme
and the fragments corresponding to the 1stH and 2ndKHS1
domains. As presented in the left lanes of Figures 5A to 5C
(lanes 1), the RNA was degraded promptly by the FL and 1stH
proteins and very slowly by the 2ndKHS1 proteins, as shown
in Figure 2. However, the product containing the stem-loop structure at the 3 end accumulated only for the FL protein. Therefore,
this result indicates that only the protein containing two RNase
PH core domains paused at a stem-loop structure. However, because the activity of the 2nd core on nonpolyadenylated RNA is
very low, we repeated the experiment using a polyadenylated
version of the same substrate. Indeed, as observed in Figure 2,
the polyadenylated molecule, unlike the nonpolyadenylated
RNA, was degraded at a similar rate by the FL, 1stH, and
2ndKHS1 proteins. Also in this case, a product produced as a
result of pausing at the stem-loop structure was detected only
2010
The Plant Cell
for the FL protein and not for either of the two fragmented proteins (Figure 5, lanes 2). As reported previously, the accumulation of this product was reduced significantly when the substrate was polyadenylated compared with the nonpolyadenylated
version (Lisitsky et al., 1996, 1997b). Because the 3 ends of most
transcripts are characterized by stem-loop structures formed by
exonucleolytic processing involving PNPase (Walter et al., 2002),
this observation suggests a functional explanation for a PNPase
enzyme harboring two RNase PH domains.
Competition of PNPase and Each of the RNase PH Core
Domains for Polyadenylated RNA
We reported previously that both the chloroplast and the E. coli
PNPases compete for polyadenylated RNAs and that the biochemical reason is their high affinity for poly(A) ( Lisitsky et al.,
1997b; Lisitsky and Schuster, 1999). This competition was observed in experiments in which two RNA molecules, one polyadenylated and the other nonpolyadenylated, were incubated
with the enzyme. When incubated separately, the degradation
rates were similar. However, when mixed and incubated with
limited amounts of enzyme, the nonpolyadenylated RNA was stable but the polyadenylated RNA was degraded rapidly ( Lisitsky et
al., 1997b; Lisitsky and Schuster, 1999). This result is interpreted
most easily as preferential binding of the enzyme to the polyadenylated RNA, and because PNPase works processively, it is sequestered from the nonpolyadenylated RNA.
In the experiments reported here, we investigated whether
the PNPase fragments also competed for polyadenylated substrates. Equal amounts of polyadenylated and nonpolyadenylated RNAs were mixed and incubated with the FL, 1stH, and
2ndKHS1 proteins. As shown in Figure 5A (lanes 3) for the FL
protein, the degradation rates for the two RNAs were similar
when incubated separately. However, when mixed, the degradation rate of the nonpolyadenylated RNA was reduced markedly. Stabilization of the nonpolyadenylated RNA also was observed with the 1stH protein, but it was impossible to determine
for the 2ndKHS1 protein, because the degradation rate of the
nonpolyadenylated RNA was very slow even when incubated
separately, as described above ( Figures 2 and 5C). However, the
results presented in Figure 5C clearly show that polyadenylated
RNA is degraded by the 2ndKHS1 protein much faster than
by the nonpolyadenylated molecule, suggesting that high poly(A)
affinity of the S1 domain is essential for efficient degradation activity. These results also suggest that the competition for polyadenylated RNA is performed by both the 1stH and 2ndKHS1
subdomains. Although the competition for polyadenylated
RNA of the 2ndKHS1 protein can be explained by the presence of the S1 domain (Figure 3), that of the 1stH protein cannot be explained by the same mechanism.
A Platform of 6 to 12 Nucleotides 3 to the Stem Loop Is
Required for RNA Polyadenylation by PNPase
Most chloroplast transcripts are characterized at the 3 end by
a stem-loop structure, and these RNA molecules are polyadenylated inefficiently by PNPase (Lisitsky et al., 1996; Schuster
et al., 1999; Monde et al., 2000). We found previously that the E.
Figure 5. Unlike the FL Protein, the Degradation Activity of Each Core
Domain Is Not Inhibited by a Stem-Loop Structure.
32P-RNA corresponding to the 3 end of the spinach chloroplast psbA
gene (296 nucleotides) (lanes 1), or the same RNA that was first elongated with 200 adenosines (lanes 2), was used as a substrate for the
PNPase and the two parts 1stH and 2ndKHS1. In lanes 3, half of the
amount of RNA molecules incubated in lanes 1 and 2 were mixed together and incubated with the proteins. Samples were withdrawn at 0,
35, 60, 90, and 120 min and analyzed by denaturing PAGE and autoradiography. Schemes of the corresponding RNA molecules are shown at
right.
coli PAP I also is inhibited by a stem-loop structure but that the
addition of two nucleotides 3 to the stem loop is sufficient to promote efficient polyadenylation (Yehudai-Resheff and Schuster,
2000). Here, we wanted to determine how many nucleotides 3 to
the stem-loop structure are required to promote polyadenylation
by the chloroplast PNPase. RNA was prepared representing the
E. coli malE-malF transcript, whose stable stem-loop structure
was strengthened further by modifying position 5 from the base
of the stem loop from A to C (Blum et al., 1999; Yehudai-Resheff
and Schuster, 2000). The same RNA having 6, 12, or 24 nucleotides 3 to the stem loop also was prepared, as shown schematically at the bottom of Figure 6. Each of the RNAs was incubated
with the FL enzyme and ADP in a polyadenylation assay. The results showed no activity without the addition of nucleotides to the
stem loop, very little polyadenylation activity with 6 nucleotides,
and significant activity with 12 nucleotides. Therefore, the spinach chloroplast PNPase requires 6 to 12 nucleotides 3 to the
stem loop for efficient polyadenylation. Together with the results
presented in Figure 5, it is evident that a stable RNA stem-loop
structure inhibits the processive degradation and polymeriza-
Domain Analysis of PNPase
Figure 6. A Tail of 6 to 12 Nucleotides at the 3 End of the Stem Loop Is
Required for Polyadenylation by PNPase.
32P-RNAs
corresponding to the 3 end of the E. coli malE-malF mRNA
with the addition of 6, 12, or 24 nucleotides 3 to the stem loop were
tested for polyadenylation using the FL PNPase and ADP. The reaction
was stopped at 0, 35, and 60 min, and the RNA was purified and analyzed by denaturing PAGE and autoradiography. Some RNA molecules
ending at the 3 end at the stem-loop structure were produced as a result of the early termination of the T7 RNA polymerase in the in vitro
transcription reaction.
tion activity of the enzyme, clarifying its known function as a
protective cis element.
Similarities and Differences between the Amino Acid
Sequences of the Two Core Domains of PNPase and
RNase PH
The observation that the chloroplast PNPase is composed of
two active domains homologous with RNase PH raised the
question of the similarity between the two domains and whether
these domains are the result of a duplication event of a common
RNase PH ancestor. Figure 7 presents a multiple sequence alignment of the two domains from several bacterial and eukaryotic
nuclear genes that encode mitochondrial and chloroplast PNPases as well as E. coli RNase PH. First, each core domain was
aligned to RNase PH of E. coli. In the second step, the two multiple alignments obtained were combined. Finally, the combined
alignment was adjusted manually to give the best identity according to the crystallographic structure of the S. antibioticus PNPase
(indicated in the first and last lines of Figure 7). Identical and similar amino acids between the two domains are indicated with dark
and bright gray backgrounds, respectively, whereas homology
restricted to the 1st or 2nd core domain is colored blue or red, respectively.
As observed previously, several regions highlight the homology of the two core domains with each other and with RNase
PH (Zuo and Deutscher, 2001; Aloy et al., 2002; Raijmakers et
al., 2002; Symmons et al., 2002). Some highly conserved regions
were observed in each of the domains. For example, the residues spanning positions 170 to 192 (which actually continue at
positions 204 to 217) are very conserved in all of the PNPase
2nd core domains and also contain the tungsten binding site
2011
and two residues in which mutations of the E. coli enzyme inhibited phosphorolysis activity ( Symmons et al., 2000; Jarrige
et al., 2002). Interestingly, four of the amino acids that eliminated
degradation activity when mutated in the E. coli PNPase were
conserved in both domains (residues 35, 124, 163, and 249)
(Jarrige et al., 2002). Of the others, the 1st core mutation at position 147 eliminated degradation activity of the enzyme and was
conserved completely in this domain but was not conserved in
the 2nd domain. However, the functionally important residue at
position 163 was conserved only in bacterial PNPases, and the
residue at position 181 was conserved only in the 2nd domain.
In addition, the amino acids that participate in the formation of
the S. antibioticus activity site, as identified by the binding of
tungsten to the protein (positions 128, 174, 175, and 176), were
conserved only in the 2nd domain, in agreement with the biochemical observation of this work that only the 2nd core is active in polymerization. Finally, it is clear from Figure 7 that the
1st core of the mitochondria enzymes is more different from its
bacterial and chloroplast counterparts but is closer to the 2nd
core consensus (e.g., positions 101, 126, 127, and 145).
The 1st domain was characterized as possessing different
activities in diverse organisms, including the synthesis of the
nucleotide ppGppp in S. antibioticus, the absence of this activity in E. coli (Symmons et al., 2000), and RNA degradation but
not polymerization in spinach chloroplasts ( Figure 2). As discussed above, the tungsten binding/phosphorolysis activity site
of the S. antibioticus enzyme was highly conserved in the 2nd
domain but shared only limited identity with the 1st core domain.
Interestingly, alignment of the 1st and 2nd core sequences to the
RNase PH disclosed that most of the amino acids conserved in
the two cores also were conserved in RNase PH (Figure 7, middle line). However, certain sequences of the RNase PH were better conserved in the 1st domain, whereas others were much
better conserved in the 2nd domain. As expected, the identity
was best when comparing related enzymes in each group. For
example, the human and mouse PNPases, which are nucleus encoded but probably targeted to mitochondria, share a very high
degree of identity ( Figure 7). At several locations, both enzymes
differ from the sequences conserved in most of the plant and
bacterial enzymes (e.g., residues 48, 50, 114 to 118, 126, 127,
and 134). Additional deletions and site-directed mutagenesis of
each of the two domains is required to define exactly the degradation, phosphorolysis, polyadenylation, and ppGppp synthesis activity sites of PNPase in different organisms.
DISCUSSION
Activities of the Different Domains
PNPase plays a pivotal role in RNA degradation and polyadenylation in the chloroplast. After the initial cleavage by an endoribonuclease, which could be an RNase E/G protein, CSP41,
or the 54 kD protein (Nickelsen and Link, 1993; Yang et al., 1996;
V. Liveanu and G. Schuster, unpublished data), the cleavage
product is modified by the addition of a poly(A)-rich tail, which
can be several hundred nucleotides long. Only then is the polymerized RNA rapidly degraded exonucleolytically by PNPase
and possibly by additional exoribonucleases such as RNase
2012
The Plant Cell
Figure 7. Multiple Sequence Alignment of the 1st and 2nd Core Domains of PNPases.
The two core domain sequences of PNPases from chloroplast (C), mitochondria (M), and bacteria (B), as well as the E. coli RNase PH (RPH), were
aligned to show modifications subsequent to the gene duplication fusion event. In addition, the known crystallographic structure of the S. antibioticus
also was applied to fine-tune the alignment. In the lines marked “structure,” the secondary structure is indicated: H represents an -helix, E represents a -strand, T represents a turn with hydrogen bonding, and G represents a 310-helix. Green shading indicates identity in the structure of the
two cores. The organisms are as follows: At, Arabidopsis thaliana; So, Spinacia oleracea; Hs, Homo sapiens; Mu, Mus musculus; Dm, Drosophila melanogaster; Ce, Caenorhabditis elegans; Ec, Escherichia coli; Pa, Pseudomonas aeruginosa; Sa, Streptomyces antibioticus; Bs, Bacillus subtilis; and
Sy, Synechocystis sp PCC6803. Amino acids were grouped by polarity as follows: (1) R and K; (2) D, E, Q, and N; (3) W, Y, and F; (4) I, V, L, M, and A;
and (5) S and T. In addition, the amino acids A, S, and G were grouped by virtue of their small sizes and are framed in black. The gaps were introduced to allow maximum alignment of the two core and RNase PH domains. Locations at which 50% of the amino acids belong to the same group
are boxed in gray. Similarities within the 1st and 2nd core domains are marked in red and blue, respectively. Locations of 75% identity are colored
with dark gray. The numerals 1 and 2 below the alignment indicate the site-directed mutations in the 1st and 2nd cores of the E. coli PNPase, respectively, resulting in the inhibition of RNA degradation activity (Jarrige et al., 2002). The letter T below the alignment at positions 128, 174, 175, and 176
indicates the tungsten binding site of the S. antibioticus PNPase (Symmons et al., 2000).
II/R. In spinach chloroplasts, both polymerization and degradation are performed by PNPase, because there is no evidence
for PAP like that found in E. coli (Yehudai-Resheff et al., 2001).
This also appears to be the situation in cyanobacteria, which is
believed to be evolutionarily related to the chloroplast ancestor
that invaded the primitive eukaryotic cell (Rott et al., 2003). However, only homogenous poly(A) tails were observed in Chla-
mydomonas reinhardtii chloroplasts, suggesting that perhaps a
PAP exists there (Komine et al., 2000). In addition, the 3 end
processing of chloroplast RNA molecules was hampered in
an Arabidopsis line in which the expression of PNPase was inhibited, indicating the importance of this enzyme for that process
(Walter et al., 2002). Interestingly, under this condition, increased
amounts of polyadenylated chloroplast RNAs were detected,
Domain Analysis of PNPase
2013
Figure 7. (continued).
suggesting that the remaining PNPase molecules are highly
shifted to the polymerization mode of activity or that other polyadenylation activity could take place when PNPase is absent or
reduced (Walter et al., 2002).
Because the PNPase conserved structure is composed of
two RNase PH–like domains, two RNA binding domains ( KH and
S1), and one -helical domain between the two RNase PH–
related domains, we decided to study the RNA degradation and
polymerization of the two RNase PH domains as well as the RNA
binding properties of each domain. To this end, recombinant
PNPase fragments were produced in an E. coli strain lacking the
endogenous PNPase and their activities were characterized.
The results are described in Table 1 and are discussed below.
The 1st Core Domain
The 1st core domain did not bind the tungsten phosphate analog in the S. antibioticus PNPase, as seen in the crystal struc-
ture; therefore, it was not considered to be the location of the
phosphorolysis catalytic domain (Symmons et al., 2000). However, one mutation of the E. coli enzyme in this domain eliminated catalytic activity, whereas several others reduced it markedly (Jarrige et al., 2002). In addition, compared with the 2nd
core domain, the amino acid sequence is less conserved for the
different species (Figure 7). Also, the PNPase activity of the synthesizing ppGppp component in S. antibioticus is believed to
be localized in the 1st core domain, whereas no such activity is
performed by the E. coli enzyme (Symmons et al., 2000). Therefore, the results presented here that this domain is active in RNA
degradation but not in polymerization were somewhat surprising. If the lack of RNA degradation activity were found for this
domain in other bacterial PNPases, it might suggest that the
phosphorylase site was converted during evolution to perform
other functions, such as ppGppp synthesis in S. antibioticus. In
addition, this site could be converted in the spinach chloroplast
PNPase to be active only in RNA degradation and not in poly-
2014
The Plant Cell
Table 1. Activities of the PNPase Domains
Activity
RNA Binding
Protein
Degradation
Polymerization
Poly(A)
Poly(U)
RNA
Complementation of E. coli pnp
FL
FL-S1
1stH2nd
1stH
2ndKHS1
2nd
KHS1
NDa
b
b
ND
ND
ND
ND
a ND,
Not determined.
degradation activity for the 2nd core domain was obtained only with polyadenylated RNA or RNA molecules that were first polymerized by the
enzyme.
b RNA
merization. Indeed, we detected no polymerization activity of the
1st core domain even with GDP under conditions in which this
activity was detected easily, presumably because of the lack of
RNA degradation activity (Figure 2). Therefore, further studies are
required to determine why the 1st core is active only in degradation and not in polymerization.
Activity of the 2nd Core Domain
The binding of the Pi analog tungsten only to the 2nd core domain in S. antibioticus PNPase, together with the greater similarity of the amino acid sequence of the 2nd core from different
PNPases, made it the obvious candidate to harbor the phosphorolytic active site (Symmons et al., 2000). In addition, most
of the mutations of the E. coli enzyme that inhibit this activity
were located in this domain (Jarrige et al., 2002). Therefore, we
were surprised initially to observe very little RNA degradation
activity by the protein constructs harboring this and not the 1st
domain (Figure 2). This discrepancy was resolved when these
proteins were supplied with polyadenylated RNAs (Figure 5) or
when nucleosides-diphosphate were added to the reaction
mixture (Figure 2). Then, the degradation of polyadenylated
RNA was at the same rate as that of the FL or the 1st core domain, and more interestingly, nonpolyadenylated RNAs were
polymerized initially and only then degraded (Figure 2).
The observation that the phosphorolytic activity site of the
spinach chloroplast PNPase is active only on polyadenylated
RNAs, and is first polymerized transiently and only then degraded when nonpolyadenylated RNA is the substrate, is similar to the results obtained from RNA degradation assays performed with lysed chloroplasts. When lysed chloroplasts were
supplied with 32P-RNA, it was initially polymerized transiently and
only then degraded (Yehudai-Resheff et al., 2001). Moreover,
similar to what was observed here for the 2nd core domain, the
initial polymerization was required for the subsequent degradation step, because the addition of GDP resulted in polyguanylated
RNA that was resistant to degradation by exoribonucleases
(Lisitsky et al., 1997a; Yehudai-Resheff et al., 2001) ( Figure 2).
Because the chloroplast tails are poly(A) rich and the S1 domain, characterized in this work to be the site of the high-affinity poly(A) binding, is located next to the 2nd core, it also probably is involved in the polymerization and degradation of the
polyadenylated RNA. However, the mechanism may not be as
simple as portrayed above, because the protein composed of
only the 2nd core domain without the S1 also displayed this behavior (Figure 2). It will be interesting to investigate whether the
2nd domain of the bacterial and mitochondria PNPases, and
possibly some of the corresponding proteins of the exosome,
share this property.
PNPase Structure
The stem-loop structures located at the 3 ends of most of the
chloroplast transcripts play an important role as cis elements
for 3 end processing, protection from exoribonuclease attack,
and the inhibition of polyadenylation that is followed by degradation (Monde et al., 2000). A well-known phenomenon of PNPase and RNase II is that they pause at sites containing doublestranded RNA, often being formed by the inverted repeats that
form stem-loop structures. Here, we have shown that only the
proteins that contain the two core domains pause at the stemloop structures. Each independent domain completely degrades
the RNA molecule without pausing at the stem-loop structure
(Figure 5). This phenomenon is best explained by the mechanism
Domain Analysis of PNPase
already suggested whereby the RNA to be degraded enters the
“hole” of the doughnut-shaped structure of the homotrimeric enzyme, which fits the size of a single-stranded RNA molecule
(Symmons et al., 2000). A double-stranded RNA cannot enter
this hole, and the enzyme is stuck when it reaches a stem-loop
structure.
Continuing with this hypothesis, it is tempting to suggest that
the formation of the doughnut-shaped structure of the homotrimer is related to the requirement of a system that should be inhibited at a stem-loop structure located at the 3 ends of the
chloroplast and perhaps prokaryotic mRNAs as well. A doughnut-shaped structure also is a characteristic of other processive enzymes. It will be interesting to analyze the biochemical
properties of the exosome complex, because it has been suggested to share a similar structure (Aloy et al., 2002; Raijmakers
et al., 2002; Symmons et al., 2002). In addition, a stem-loop
structure is elongated inefficiently by PNPase and the E. coli
PAP (Yehudai-Resheff and Schuster, 2000; Yehudai-Resheff et
al., 2001; this work). A tail of 6 to 12 nucleotides 3 to the stem
is required for PNPase to polyadenylate the RNA ( Figure 6). There-
2015
fore, aside from protecting the transcript by inhibiting degradation
activity, the stem-loop structure also protects the RNA by inhibiting
polyadenylation.
Evolution of the PNPase and the Exosome
The multiple sequence alignment of the PNPases from bacteria
and organelles, as well as the exosome proteins, enabled the
creation of a phylogenetic tree that clearly revealed four separate branches (excluding the single line of the E. coli RNase PH)
(Figure 8). On the prokaryotic side, the duplicated ancient RNase
PH domains remained linked in the same polypeptide, resulting
in a single protein containing the two core domains followed by
the two RNA binding domains. Three such proteins then could
form the homotrimeric structure containing six core domains
and three of each of the RNA binding KH and S1 domains. Interestingly, the first gene duplication that probably formed the
PNPase ancestor occurred only once during evolution, and the
PNPase proteins found today in the bacteria and organelles of
all organisms originated from this event. The chloroplast en-
Figure 8. Phylogenetic Tree of the RNase PH Domains of Bacterial and Organelle PNPases and Exosome Proteins.
The 1st and 2nd core domains of PNPases, the related exosome proteins, and the E. coli RNase PH were aligned using the CLUSTAL X multiple sequence alignment tool, and a phylogenetic tree was constructed as described in Methods. Proteins from the same organism are colored alike. The
dotted lines indicate regions of the tree where the bootstrap value was 50%; therefore, the validity of these lines is low. The organisms are as follows: At, Arabidopsis thaliana; So, Spinacia oleracea; Hs, Homo sapiens; Ec, Escherichia coli; Sa, Streptomyces antibioticus; Sy, Synechocystis sp
PCC6803; St, Staphylococcus aureus; Dm, Drosophila melanogaster; Ce, Caenorhabditis elegans; Ye, Saccharomyces cerevisiae; and Mt, Methanobacterium thermoautotrophicum. In M. thermoautotrophicum, the proteins homologous with Rrp43 and Rrp41 are MTH682 and MTH683, respectively. Accession numbers are given in Table 3.
2016
The Plant Cell
zymes are related more closely to each other than to the bacterial and mitochondria enzymes; therefore, they were resolved to
different branches. The exosome structure was described recently as very similar to that of the PNPase trimer, as shown in
Figure 1, suggesting functional and structural reasons for the
“six-domain structure” composed of three PNPases or 10 to 11
exosome proteins (Aloy et al., 2002; Raijmakers et al., 2002;
Symmons et al., 2002). The results of this analysis suggest close
evolutionary relationships between the bacterial and organelle
PNPases and the proteins of the exosome.
The related exosome proteins also were divided clearly into
two completely separate branches, suggesting that each branch
was either derived from one of the PNPase core domains or, alternatively, functionally related to it (Figure 8). Therefore, it is
tempting to relate each branch to one of the PNPase core domains. However, the sequence alignment did not reveal a definite relationship of each branch to one of the PNPase core domains. Indeed, in previous alignments, Rrp42 was assigned to
the 1st (Raijmakers et al., 2002) or 2nd (Aloy et al., 2002) core domain. A similar situation was described for PMScl75/Rrp45 (Aloy
et al., 2002; Raijmakers et al., 2002). Here, the multiple alignment
analysis and the derived phylogenetic tree, as presented in Figure 8, did not significantly favor the assignment of each branch
of the exosome proteins to one of the PNPase core domains.
As with the PNPase branches, there is a complete separation
between these two exosome proteins after the first duplication
of the ancient RNase PH domain in all species, including yeast,
plants, and mammals. Of evolutionary interest is the Methanococcus lineage of archaea, in which these proteins, although
separated in the phylogenetic tree (as are the yeast, plant, and
mammalian exosomes), are located on the same operon and
possibly are derived from the same primary transcript ( Koonin
et al., 2001). It will be interesting to analyze the exosome of Methanococcus in light of the sequence data obtained to date from
the Halobacterium lineage, in which no exosome proteins or
PNPase were detected (Koonin et al., 2001). It also is interesting that although an exosome is present in yeast, no genomic
or biochemical evidence for a mitochondrial PNPase was obtained (Figure 8). On the other hand, a complex composed of
an RNase II–like ribonuclease and an RNA helicase was found
( Dziembowski et al., 2002). By contrast, mitochondrial PNPases and RNA polyadenylation are present in mammals and
plants (Figure 8) (Ojala et al., 1981; Gagliardi and Leaver, 1999;
Lupold et al., 1999). Therefore, the Halobacterium lineage and
Saccharomyces cerevisiae mitochondria might represent systems in which RNA degradation is dependent exclusively on a
hydrolytic RNase II–like enzyme, perhaps without any phosphorylase or RNA polyadenylation involved (Dziembowski et al.,
2002).
The recent analysis of the bacterial and organelle PNPases
compared with the exosome proteins, together with the structural analysis of the two protein complexes, revealed a theme
generally conserved in bacteria, chloroplasts, mitochondria, the
cytoplasm, and the nucleus and possibly also in some archaea.
In all of these systems, it probably is engaged in RNA degradation and processing. Analysis of the biochemical properties
of each domain of the spinach chloroplast PNPase revealed
unique features that may be related to the general function of
this RNA degradation machine or specifically to the spinach chloroplast enzyme. For example, both complexes degrade RNA molecules containing poly(A) tails and display high affinity for A- and
U-rich sequences (Mukherjee et al., 2002; this work). The molecular and biochemical analysis of additional PNPases and exosomerelated proteins, as well as an additional analysis of the spinach
PNPase, will reveal the molecular details of the mechanism of action of this evolutionarily conserved RNA degradation machine.
METHODS
Production of Recombinant PNPase and Its Fragmented Versions
The corresponding DNA sequences of the mature protein (without the
transit peptide) and the different fragmented versions were amplified by
PCR using the primers listed in Table 2, and spinach (Spinacia oleracea)
oligo(dT)-primed cDNA was prepared as described previously (Baginsky et
al., 2001). For expression in Escherichia coli, the PCR products were inserted into the Pet 20b vector (Novagen, Madison, WI) with the addition
of a His6 tag to the C terminus (Figures 1E and 1F). Because the addition
of the His6 tag to the N terminus of the E. coli PNPase had been reported
to hamper its activity (Blum et al., 1999), we added this tag at the various
C termini (Figure 1E). Nevertheless, we were unable to produce the FL
protein (lacking only the transit peptide) with the His6 tag in a soluble
form (Figure 1E). Therefore, this protein was expressed in the PT7-7 system (Citovsky et al., 1990) without the His6 tag and purified biochemically using a series of size-exclusion, heparin, and anion-exchange columns, resulting in a purified protein with a very low yield (Figure 1E).
Fortunately, we were able to produce the other parts with the His6 tag,
enabling rapid and easy purification, and at high yields (Figure 1F).
An additional obstacle in the purification of the overexpressed proteins in E. coli was that the recombinant proteins copurified with the E.
coli PNPase because of the formation of heterotrimers between the
chloroplast and bacterial subunits (data not shown). Use of the PNPaseless strain ENS134-3 containing the T7 RNA polymerase [BL21(DE3)
(lacZ::Tn10 malPp
534::PT7lacZ-Arg5)(pnp::Tn5)] (Lopez et al., 1999), in
which the E. coli PNPase is not expressed because of the insertion of the
Tn5 transposable element (kindly obtained from Marc Dreyfus, Ecole
Normale Supérieure, Paris, France), resolved this problem. Expression
and purification were performed according to the manufacturer’s protocol using a nitrilotriacetic acid agarose affinity column with an additional
purification step using a MonoQ column (Pharmacia). The FL protein was
expressed in the same cells, and the recombinant protein was purified
biochemically using heparin and size-exclusion Superdex 200 and
MonoQ columns (Pharmacia). All of the proteins were purified to one
SDS-PAGE silver-stained band (Figure 1F) without any contamination
activity of other ribonucleases detected.
Synthetic RNAs
The plasmids used for the in vitro transcription of parts of the spinach
chloroplast psbA and petD-Dra were described previously (YehudaiResheff et al., 2001). The E. coli malE-malF intergenic region, which contained a stable stem loop in which the nucleotide at position 5 from
the base of the stem loop was modified from A to C to ensure the stabilization of the stem-loop structure, also was described previously
(Yehudai-Resheff and Schuster, 2000). RNAs were transcribed using T7
RNA polymerase and radioactively labeled with -32P–UTP, and the FL
transcription products were purified from 5% denaturing polyacrylamide
gels (Lisitsky et al., 1996). For the preparation of transcripts terminated
with the poly(A) sequence, the psbA transcription product was incubated with the yeast poly(A)-polymerase (Pharmacia) and 1 mM ATP for
Domain Analysis of PNPase
2017
Table 2. Oligonucleotides Primers Used for the Expression of PNPase Fragmented Proteins
Protein
Name
Sequence
FL
F-172
R-2466
F-172
R-2220
F-172
R-1890
F-172
R-1185
F-1227
R-2400
F-1227
R-1890
F-1891
R-2400
F
R
5-GGAATTCGCATATGGTTAGAGCTATGGCTCAA-3
5-CGGGATCCCCAAGCACGCCGACTAAGGC-3
FL-S1
1stH2nd
1stH
2ndKHS1
2nd
KHS1
E. coli PNPase
5-CGGGATCCCCGGCCTTCGATTTCTCAAG-3
5-CGGGATCCCCACAATCAGCATTTCCTGC-3
5-CGGGATCCCCATCACCTTCATCAACTTCGCC-3
5-GGAATTCCATATGTTCTCTGAGGTAGATGTG-3
5-CGGGATCCCCTCCAACTTTAAATGCAT-3
5-GGAATTCCATATGGTTACAGCATTCCAAATG-3
5-GGAATTCGCATATGATGCGCAGAAGATCGGGTAT-3
5-GCCCAAGCTTCTCGCCCTGTTCAGCAGCCG-3
Restriction sites in the primer sequence are underlined (NdeI and BamHI). The name of the primer designates the direction (F, forward; R, reverse) and
the number of the first nucleotide counting from the adenosine of the first ATG. If the primer was used several times, the nucleotide sequence is
shown only once.
30 min, and the elongated RNA was purified by denaturing PAGE
(Lisitsky et al., 1996).
RNA Polyadenylation and Degradation Assays
Polyadenylation and degradation activities of the recombinant proteins
were assayed as described previously (Yehudai-Resheff et al., 2001).
Briefly, 32P-RNA (1 fmol) was incubated with the corresponding protein
(1 fmol) in buffer E (20 mM Hepes, pH 7.9, 60 mM KCl, 12.5 mM MgCl2,
0.1 mM EDTA, 2 mM DTT, and 17% glycerol) at 25C for the times indicated in the figures. When polyadenylation or polyguanylation were assayed, the corresponding nucleotide was added, also as indicated in the
figures. For the degradation assay, Pi was added as indicated in the figures. After incubation, the RNA was isolated and analyzed by denaturing
PAGE and autoradiography. For thin layer chromatography analysis of
the degradation products, an aliquot of each sample was spotted on a
polyethyleneimine thin layer chromatography plate, which then was developed with 1 M LiCl, dried, and exposed to autoradiography. Nucleosides mono-, di-, and three phosphates (5 g of each) were separated
on the same plate and visualized by fluorescence quenching.
UV Light Cross-Linking
UV light cross-linking of proteins to radiolabeled RNA was performed as
described previously (Lisitsky et al., 1997b). The proteins (10 fmol) were
mixed with 32P-RNA (10 fmol) in the buffer containing 10 mM HepesNaOH, pH 7.9, 30 mM KCl, 6 mM MgCl2, 0.05 mM EDTA, 2 mM DTT, and
8% glycerol and cross-linked immediately with 1.8 J of UV irradiation in
a UV light cross-linker (Hoefer, San Francisco, CA). This was followed by
digestion of the RNA with 10 g of RNase A and 30 units of RNase T1 at
37C for 1 h. The proteins then were fractionated by SDS-PAGE and analyzed by autoradiography. For the competition assay, the protein was
mixed with ribohomopolymers for 5 min, and the radiolabeled RNA was
added. An average length of 400 nucleotides was used to calculate the
molar amount.
Complementation of the Growth of E. coli PNPase-less Mutants
The SK 8992 strain (thyA715, l-, rph-1, rna-19, pnp::Tn5, KanR), in which
the gene encoding PNPase was inactivated by the insertion of the transposable element Tn5 and cannot grow at 18C, was obtained from Sidney Kushner (Athens, GA). This strain differs from the ENS134 strain
(pnp::Tn5) used for the expression of recombinant proteins because it is
not a derivative of BL21(DE3) and therefore lacks the gene for the T7
RNA polymerase. We had to use this strain in the complementation experiments because we found that the insertion of any plasmid containing
the T7 promoter sequence to the ENS134 strain resulted in the complementation of growth at 18C (data not shown). In addition, this strain also
lacks the other Pi-dependent exoribonuclease, RNase PH. Changing the
E. coli strain to SK 8992 lacking the T7 RNA polymerase solved this
problem. The different deleted PNPase constructs were cloned into the
PT7-7 plasmid (Citovsky et al., 1990), transformed into this strain, and
spotted on Luria-Bertani agar plates containing 100 g/mL ampicillin
and 25 g/mL kanamycin. Because of the lack of the T7 RNA polymerase in this strain, it is unclear how abundantly the recombinant proteins
were expressed. However, immunoblot analysis of protein extracts of
the transformed bacteria clearly defined the accumulation of the recombinant proteins (Figure 4). As a positive control, the E. coli PNPase was
amplified by PCR from genomic DNA using the primers listed in Table 1
and introduced into the same plasmid. Incubation time was 16 h at 37C
and 48 h at 18C. Immunoblot analysis of the bacteria grown at either 37
or 18C was used to follow the production of the corresponding proteins.
Structure Prediction, Complex Model Building,
and Sequence Analysis
Homology-based modeling of the three-dimensional structure of the spinach PNPase was performed using the 3D-PSSM Fold Recognition Server
at http://www.sbg.bio.ic.ac.uk/~3dpssm/. The visualization of the spinach PNPase trimer complex was performed by inserting the monomeric
model described above into a pseudo rhombohedral (H32) space group
with dimensions similar to those of the Streptomyces antibioticus crystal
structure (PDB code 1E3H) using Quanta (Accelrys, San Diego, CA).
2018
The Plant Cell
Table 3. Exosome Proteins That Display Homology with the Core
Domains of PNPase
Organism
Symbol
Protein
Accession No.
Arabidopsis thaliana
At
Rrp41
Rrp42
Rrp43
PM/Scl-75
Rrp46
Rrp41
Rrp42
Rrp43
PM/Scl-75
Rrp46
Rrp41
Rrp42
PM/Scl-75
Rrp46
Rrp41
Rrp42
PM/Scl-75
Rrp46
Ski6
Rrp42
Rrp45
Rrp46
MTH682
MTH683
NP_191721
AAF13093
NP_176216
NP_566441
NP_190207
Q9NPD3
Q15024
Q96B26
Q06265
Q9NQT4
Q17533
NP_508024
T28842
NP_496284
AAF53263
AAF58076
AAF48665
AAF54530
NP_011711
NP_010172
NP_010566
NP_011609
H69190
NP_275826
Homo sapiens
Hs
Caenorhabditis elegans
Ce
Drosophila melanogaster
Dm
Saccharomyces cerevisiae
Ye
Methanobacterium
thermoautotrophicum
Mt
Sequences of proteins homologous with one of the core domains of the
spinach chloroplast PNPase were obtained from the National Center for
Biotechnology Information database by Basic Local Alignment Search
Tool (BLAST) search.
Sequence alignment of the different PNPases and the exosome components was performed by multiple alignment using CLUSTAL X and
motif search for all of the proteins together and finally by manually improving the alignment. The phylogenetic tree was built using the neighbor-joining method with bootstrap (CLUSTAL X). Then, to improve the
validity of the tree, the parsimony method with bootstrap (PAUP) was
used.
Upon request, materials integral to the findings presented in this publication will be made available in a timely manner to all investigators on
similar terms for noncommercial research purposes. To obtain materials,
please contact G. Schuster, [email protected].
Accession Numbers
The accession numbers of the different PNPase proteins shown in Figure
7 are as follows: At, Arabidopsis thaliana (NP_187021 chloroplast,
NP_196962 mitochondria); So, Spinacia oleracea (AAC49669); Hs, Homo
sapiens (XP_048088); Mu, Mus musculus (BAB23374); Ec, Escherichia
coli (P05055); Pa, Pseudomonas aeruginosa (AAG08126); Sa, Streptomyces antibioticus (AAB17498); Bs, Bacillus subtilis (NP_657775); and Sy,
Synechocystis sp PCC6803 (BAA16661).
ACKNOWLEDGMENTS
We thank Agamemnon J. Carpousis, Stanley Cohen, Marc Dreyfus, Sydney
Kushner, and Philippe Regnier for the E. coli strains as well as for numerous discussions, encouragement, and valuable advice. We thank Varda
Liveanu for her many suggestions and critical reading of the manuscript,
Oded Beja for helping with the phylogenetic analysis, and David Stern
for help in editing the manuscript. This work was supported by grants
from the Israel Science Foundation and the Israel-U.S. Binational Agriculture Research and Development Foundation.
Received April 28, 2003; accepted June 25, 2003.
REFERENCES
Aloy, P., Ciccarelli, F.D., Leutwein, C., Gavin, A.C., Superti-Furga,
G., Bork, P., Bottcher, B., and Russell, R.B. (2002). A complex prediction: Three-dimensional model of the yeast exosome. EMBO Rep.
3, 628–635.
Baginsky, S., Shteiman-Kotler, A., Liveanu, V., Yehudai-Resheff, S.,
Bellaoui, M., Settlage, R.E., Shabanowitz, J., Hunt, D.F., Schuster,
G., and Gruissem, W. (2001). Chloroplast PNPase exists as a homomultimer enzyme complex that is distinct from the Escherichia coli
degradosome. RNA 7, 1464–1475.
Beran, R.K., and Simons, R.W. (2001). Cold-temperature induction of
Escherichia coli polynucleotide phosphorylase occurs by reversal of
its autoregulation. Mol. Microbiol. 39, 112–125.
Blum, E., Carpousis, A.J., and Higgins, C.F. (1999). Polyadenylation
promotes degradation of 3-structured RNA by the Escherichia coli
mRNA degradosome in vitro. J. Biol. Chem. 274, 4009–4016.
Carpousis, A.J., Vanzo, N.F., and Raynal, L.C. (1999). mRNA degradation,
a tale of poly(A) and multiprotein machines. Trends Genet. 15, 24–28.
Citovsky, V., Knorr, D., Schuster, G., and Zambryski, P. (1990). The
P30 movement protein of tobacco mosaic virus is a single-strand
nucleic acid binding protein. Cell 60, 637–647.
Clements, M.O., Eriksson, S., Thompson, A., Lucchini, S., Hinton,
J.C., Normark, S., and Rhen, M. (2002). Polynucleotide phosphorylase is a global regulator of virulence and persistency in Salmonella
enterica. Proc. Natl. Acad. Sci. USA 99, 8784–8789.
Coburn, G.A., and Mackie, G.A. (1999). Degradation of mRNA in Escherichia coli: An old problem with some new twists. Prog. Nucleic
Acids Res. 62, 55–108.
Craven, M.G., Henner, D.J., Alessi, D., Schauer, A.T., Ost, K.A.,
Deutscher, M.P., and Friedman, D.I. (1992). Identification of the rph
(RNase PH) gene of Bacillus subtilis: Evidence for suppression of coldsensitive mutations in Escherichia coli. J. Bacteriol. 174, 4727–4735.
Dziembowski, A., Piwowarski, J., Hoser, R., Minczuk, M.,
Dmochowska, A., Siep, M., Van Der Spek, H., Grivell, L., and Stepien,
P.P. (2002). The yeast mitochondrial degradosome: Its composition,
interplay between RNA helicase and RNase activities and the role in
mitochondrial RNA metabolism. J. Biol. Chem. 278, 1603–1611.
Gagliardi, D., and Leaver, C.J. (1999). Polyadenylation accelerates the
degradation of the mitochondrial mRNA associated with cytoplasmic
male sterility in sunflower. EMBO J. 18, 3757–3766.
Grunberg-Manago, M. (1999). Messenger RNA stability and its role in
control of gene expression in bacteria and phages. Annu. Rev. Genet.
33, 193–227.
Hayes, R., Kudla, J., and Gruissem, W. (1999). Degrading chloroplast
mRNA: The role of polyadenylation. Trends Biochem. Sci. 24, 199–202.
Hayes, R., Kudla, J., Schuster, G., Gabay, L., Maliga, P., and
Gruissem, W. (1996). Chloroplast mRNA 3-end processing by a high
molecular weight protein complex is regulated by nuclear encoded
RNA binding proteins. EMBO J. 15, 1132–1141.
Jarrige, A., Brechemier-Baey, D., Mathy, N., Duche, O., and Portier,
C. (2002). Mutational analysis of polynucleotide phosphorylase from
Escherichia coli. J. Mol. Biol. 321, 397–409.
Domain Analysis of PNPase
Komine, Y., Kwong, L., Anguera, M., Schuster, S., and Stern, D.B.
(2000). Polyadenylation of three classes of chloroplast RNA in
Chlamydomonas reinhardtii. RNA 6, 598–607.
Koonin, E.V., Wolf, Y.I., and Aravind, L. (2001). Prediction of the
archaeal exosome and its connections with the proteasome and the
translation and transcription machineries by a comparative-genomic
approach. Genome Res. 11, 240–252.
Kuhn, J., Tengler, U., and Binder, S. (2001). Transcript lifetime is balanced between stabilizing stem-loop structures and degradation-promoting polyadenylation in plant mitochondria. Mol. Cell. Biol. 21,
731–742.
Leszczyniecka, M., Kang, D.C., Sarkar, D., Su, Z.Z., Holmes, M.,
Valerie, K., and Fisher, P.B. (2002). Identification and cloning of
human polynucleotide phosphorylase, hPNPase old-35, in the context of terminal differentiation and cellular senescence. Proc. Natl.
Acad. Sci. USA 99, 16636–16641.
Liou, G.G., Chang, H.Y., Lin, C.S., and Lin-Chao, S. (2002). DEAD box
RhlB RNA helicase physically associates with exoribonuclease
PNPase to degrade double-stranded RNA independent of the degradosome-assembling region of RNase E. J. Biol. Chem. 277, 41157–
41162.
Lisitsky, I., Klaff, P., and Schuster, G. (1996). Addition of poly(A)-rich
sequences to endonucleolytic cleavage sites in the degradation of
spinach chloroplast mRNA. Proc. Natl. Acad. Sci. USA 93, 13398–
13403.
Lisitsky, I., Klaff, P., and Schuster, G. (1997a). Blocking polyadenylation of mRNA in the chloroplast inhibits its degradation. Plant J. 12,
1173–1178.
Lisitsky, I., Kotler, A., and Schuster, G. (1997b). The mechanism of
preferential degradation of polyadenylated RNA in the chloroplast:
The exoribonuclease 100RNP/PNPase displays high binding affinity
for poly(A) sequence. J. Biol. Chem. 272, 17648–17653.
Lisitsky, I., Liveanu, V., and Schuster, G. (1994). RNA-binding activities of the different domains of a spinach chloroplast ribonucleoprotein. Nucleic Acids Res. 22, 4719–4724.
Lisitsky, I., and Schuster, G. (1999). Preferential degradation of polyadenylated and polyuridinylated RNAs by bacterial exoribonuclease polynucleotide phosphorylase (PNPase). Eur. J. Biochem. 261, 468–474.
Littauer, U.Z., and Grunberg-Manago, M. (1999). Polynucleotide
phosphorylase. In The Encyclopedia of Molecular Biology, T.E.
Creighton, ed (New York: John Wiley & Sons), pp. 1911–1918.
Lopez, P.J., Marchand, I., Joyce, S.A., and Dreyfus, M. (1999). The
C-terminal half of RNase E, which organizes the Escherichia coli
degradosome, participates in mRNA degradation but not rRNA processing in vivo. Mol. Microbiol. 33, 188–199.
Lupold, D.S., Caoile, A.G.F.S., and Stern, D.B. (1999). Polyadenylation
occurs at multiple sites in maize mitochondrial cox2 mRNA and is
independent of editing status. Plant Cell 11, 1565–1578.
Mohanty, B.K., and Kushner, S.R. (2000). Polynucleotide phosphorylase functions both as a 3 to 5 exonuclease and a poly(A) polymerase in Escherichia coli. Proc. Natl. Acad. Sci. USA 97, 11966–11971.
Monde, R.A., Schuster, G., and Stern, D.B. (2000). Processing and
degradation of chloroplast mRNA. Biochimie 82, 573–582.
2019
Mukherjee, D., Gao, M., O’Connor, J.P., Raijmakers, R., Pruijn, G.,
Lutz, C.S., and Wilusz, J. (2002). The mammalian exosome mediates
the efficient degradation of mRNAs that contain AU-rich elements.
EMBO J. 21, 165–174.
Nickelsen, J., and Link, G. (1993). The 54 kDa RNA-binding protein
from mustard chloroplasts mediates endonucleolytic transcript 3 end
formation in vitro. Plant J. 3, 537–544.
Ojala, D., Montoya, J., and Attardi, G. (1981). tRNA punctuation model
of RNA processing in human mitochondria. Nature 290, 470–474.
Raijmakers, R., Egberts, W.V., van Venrooij, W.J., and Pruijn, G.J.
(2002). Protein-protein interactions between human exosome components support the assembly of RNase PH-type subunits into a sixmembered PNPase-like ring. J. Mol. Biol. 323, 653–663.
Regnier, P., and Arraiano, C.M. (2000). Degradation of mRNA in bacteria: Emergence of ubiquitous features. Bioessays 22, 235–244.
Rott, R., Zipor, G., Portnoy, V., Liveanu, V., and Schuster, G. (2003).
RNA polyadenylation and degradation in cyanobacteria are similar to
the chloroplast but different from Escherichia coli. J. Biol. Chem. 278,
15771–15777.
Sarkar, N. (1997). Polyadenylation of mRNA in prokaryotes. Annu. Rev.
Biochem. 66, 173–197.
Schuster, G., Lisitsky, I., and Klaff, P. (1999). Update on chloroplast
molecular biology: Polyadenylation and degradation of mRNA in the
chloroplast. Plant Physiol. 120, 937–944.
Sundquist, W.I. (1993). Conducting the G-quartet. Curr. Biol 3, 893–895.
Symmons, M.F., Jones, G.H., and Luisi, B.F. (2000). A duplicated fold
is the structural basis for polynucleotide phosphorylase catalytic
activity, processivity, and regulation. Structure 8, 1215–1226.
Symmons, M.F., Williams, M.G., Luisi, B.F., Jones, G.H., and
Carpousis, A.J. (2002). Running rings around RNA: A superfamily of
phosphate-dependent RNases. Trends Biochem. Sci. 27, 11–18.
Walter, M., Kilian, J., and Kudla, J. (2002). PNPase activity determines the
efficiency of mRNA 3-end processing, the degradation of tRNA and the
extent of polyadenylation in chloroplasts. EMBO J. 21, 6905–6914.
Yancey, S.D., and Kushner, S.R. (1990). Isolation and characterization
of a new temperature-sensitive polynucleotide phosphorylase mutation in Escherichia coli K-12. Biochimie 72, 835–843.
Yang, J., Schuster, G., and Stern, D.B. (1996). CSP41, a sequencespecific chloroplast mRNA binding protein, is an endoribonuclease.
Plant Cell 8, 1409–1420.
Yehudai-Resheff, S., Hirsh, M., and Schuster, G. (2001). Polynucleotide phosphorylase functions as both an exonuclease and a poly(A)
polymerase in spinach chloroplasts. Mol. Cell. Biol. 21, 5408–5416.
Yehudai-Resheff, S., and Schuster, G. (2000). Characterization of the
E. coli poly(A)-polymerase: Specificity to nucleotides, RNA-binding
affinities and RNA-structure dependence activity. Nucleic Acids Res.
28, 1139–1144.
Zhou, Z., and Deutscher, M.P. (1997). An essential function for the
phosphate-dependent exoribonucleases RNase PH and polynucleotide phosphorylase. J. Bacteriol. 179, 4391–4395.
Zuo, Y., and Deutscher, M.P. (2001). Exoribonuclease superfamilies:
Structural analysis and phylogenetic distribution. Nucleic Acids Res.
29, 1017–1026.